Hi there,

we are attempting to submit apbs calculation on a cluster by using queue manager. Basically we have n-frames from a trajectory and for each frame we submit an apbs calculation with torque/maui (qsub command) manager.

The calculations are very slow and I presume because it appears that the calculation is going to require more memory than is available on the single node. Each single node takes 8 apbs runs but there is only 2gb of ram per node! Therefore, some apbs runs fail (seg fault) for exceed of memory. I was thinking to compile apbs versus mpi to run in parallel fashion the calculation, but does it will help? How can we deal with this problem?

I read http://www.poissonboltzmann.org/apbs/examples/parallel-electrostatics-calculations-actin-dimer but in our case the system is small (60 residues) and the problem is that we are running distributed jobs of apbs.

any help?

thanks in advance

andrea

------=_Part_1143_13390435.1381306703126--
we are attempting to submit apbs calculation on a cluster by using queue manager. Basically we have n-frames from a trajectory and for each frame we submit an apbs calculation with torque/maui (qsub command) manager.

The calculations are very slow and I presume because it appears that the calculation is going to require more memory than is available on the single node. Each single node takes 8 apbs runs but there is only 2gb of ram per node! Therefore, some apbs runs fail (seg fault) for exceed of memory. I was thinking to compile apbs versus mpi to run in parallel fashion the calculation, but does it will help? How can we deal with this problem?

I read http://www.poissonboltzmann.org/apbs/examples/parallel-electrostatics-calculations-actin-dimer but in our case the system is small (60 residues) and the problem is that we are running distributed jobs of apbs.

any help?

thanks in advance

andrea